INFORMS Nashville – 2016
244
TA33
203B-MCC
Queueing Models
Contributed Session
Chair: Pedro Cesar Lopes Gerum, PhD Student, Rutgers University, 96
Frelinghuysen Road, CoRE Building, Room 201, piscataway, NJ, 8854,
United States,
pedro.gerum@rutgers.edu1 - Mean Value Analysis Of Mixed Queuing Networks
Ivo Adan, Eindhoven University of Technology, Den Dolech 2,
Eindhoven, 5600 MB, Netherlands,
I.Adan@tue.nl,
Vidyadhar Kulkarni
We study a mixed queuing network with multi-server stations. The mixed
network has both closed and open network components: it has a fixed number of
customers (called permanent customers) that circulate among the service stations
indefinitely, and it also serves customers (called transient customers) who enter
from outside, visit the stations in a random order and leave. We develop novel
mean-value equations for recursively computing the mean queue lengths and
mean waiting waiting times, and we study the asymptotic behavior of these
quantities in the presence of multiple bottle-neck stations as the number of
permanent customers tends to infinity.
2 - Simple And Efficient Ways For Discrete GI/G/1 Queues
Winfried K Grassmann, Professor emeritus, University of
Saskatchewan, 110 Science Place, 176 Thorvaldson Building,
Saskatoon, SK, S7N 5C9, Canada,
grassman@cs.usask.caWe present a number of simple and not so simple methods to find the distribution
of the number of elements in a GI/G/1 queue and related problems. As it turns
out, many methods described in literature are mathematically challenging, but
this does not imply that they are numerically efficient numerically. In fact, it is
our experience that the simpler methods tend to be the most efficient ones, while
also easy to understand. This leads to the suspicion that criteria for publication
typically favor mathematical elegance over practical usefulness.
3 - A Queueing System With On-demand Servers: Local Stability
Of Fluid Limits
Lam M Nguyen, PhD Student, Lehigh University, 200 West Packer
Avenue, Room 362, Bethlehem, PA, 18015, United States,
lmn214@lehigh.edu, Alexander Stolyar
We consider a system, where a random flow of customers is served by agents
invited on-demand. Each invited agent arrives into the system after a random
time, and leaves it with some probability after each service completion.
Customers and/or agents may be impatient. The objective is to design a real-time
adaptive invitation scheme that minimizes customer and agent waiting times. We
consider a queue-length-based feedback scheme; study it in the asymptotic
regime where the customer arrival rate goes to infinity; and derive a variety of
sufficient conditions for the system local stability at the desired equilibrium point.
Under these conditions, simulations show good overall performance of the
scheme.
4 - Traffic Density Analytical Model Validation And Applications
Pedro Cesar Lopes Gerum, PhD Student, Rutgers University, 96
Frelinghuysen Road, CoRE Building, Room 201, piscataway, NJ,
08854, United States,
pedro.gerum@rutgers.edu, Melike Baykal-
Gursoy, Marcelo Ricardo Figueroa
This paper compares a general equation for the probability generating function of
density for a general road system, discovered by W. Xiao and Baykal-Gursoy, with
real data from Milwaukee, Wisconsin. Furthermore, once shown the analytical
model is valid, this paper presents some insights in possible applications taken
from these formulations. These insights include improving efficiency of
evacuation in extreme scenarios, such as flooding or other weather conditions;
providing useful information to decision-makers on how to better invest their
money in infrastructure; allowing the end-user of a routing system to choose
between routes according to the risk of delay he is willing to take.
TA34
204-MCC
Provider Staffing and its Impact on Patient Flow
Sponsored: Manufacturing & Service Oper Mgmt, Healthcare
Operations
Sponsored Session
Chair: Retsef Levi, MIT, 100 Main Street, Building E62-562, Cambridge,
MA, 02142, United States,
retsef@mit.eduCo-Chair: Cecilia Zenteno, Massachusetts General Hospital,
55 Fruit Street, White 400, Boston, MA, 02114, United States,
azentenolangle@mgh.harvard.edu1 - Ed Physician Staffing Via Multi-stage Multi-class Network
Caglar Caglayan, Georgia Institute of Technology, Atlanta, GA,
30318, United States,
ccaglayan6@gatech.edu, Mustafa Y Sir,
Kalyan Pasupathy, Turgay Ayer, Yunan Liu
We propose an “intuitive”, “realistic” and “tractable” model of the emergency
department (ED) by a multi-class multi-stage queuing network with multiple
targeted service levels. Based on infinite-server approximation and offered load
analysis, we employ a modified version of square-root safety principle to
determine the right number of physicians in the ED. Our model is detailed
enough to capture the key dynamics of the ED but simple enough to understand,
infer results and implement in a clinical setting.
2 - Discrete Event Simulation Of Outpatient Flow In A
Phlebotomy Clinic
Elizabeth Olin, University of Michigan, 1205 Beal, Ann Arbor, MI,
48109, United States,
genehkim@umich.edu, Amy Cohn,
Ajaay Chandrasekaran
The University of Michigan Comprehensive Cancer Center handles approximately
97,000 outpatient visits annually, with most including a blood draw, clinic
appointment, preparation of infusion drugs, and an infusion appointment. The
goal of our project is to reduce patient waiting times at the phlebotomy (blood
draw) clinic, which appears to be a primary bottleneck in the patient experience.
In order to accomplish this goal, we developed a discrete-event simulation of the
clinic’s patient and work flow. By adjusting the various simulation parameters, we
can evaluate alternative methods to improve turnaround time, patient wait time,
and phlebotomist utilization.
3 - Quantifying Provider’s Schedule Effects On
Patient’s Length-of-Stay
Kimia Ghobadi, MIT, Cambridge, MA, United States,
kimiag@mit.edu,Andrew Johnston, Retsef Levi, Walter O’Donnell
We identify a natural randomized control setting between providers’ schedule and
patients arrival in a congested Department of Medicine in a large academic
hospital. We use this setting to build a predictive model and quantify the impact
of care team handoff on patients’ length-of-stay.
TA35
205A-MCC
Online Services: Learning and Pricing
Sponsored: Manufacturing & Service Oper Mgmt, Service
Operations
Sponsored Session
Chair: Yash Kanoria, Columbia University, Graduate School of
Business, New York, NY, 10027, United States,
ykanoria@columbia.eduCo-Chair: Vijay Kamble, Stanford University, Stanford, CA, 9, United
States,
vijaykamble.iitkgp@gmail.com1 - Efficiency And Performance Guarantees For Network Revenue
Management Problems With Customer Choice
David Simchi-Levi, Massachusetts Institute of Technology, Dept of
Civil and Environmental Engineering, 77 Massachusetts Avenue
Rm 1-171, Cambridge, MA, 02139, United States,
dslevi@mit.edu,Wang Chi Cheung
We consider the network revenue management problem with customer choice.
While the solution to the Choice-based Deterministic Linear Program (CDLP) can
be used to design a near-optimal policy, CDLP has an exponential size. We
propose algorithms that solves CDLP with polynomially many elementary
operations and invocations to an oracle that solves the underlying single period
problems. Next, we design an efficient online algorithm for the problem with
MNL choice models, where the parameters are unknown. The algorithm achieves
a regret of O(T2/3), where T is the length of the time horizon.
2 - Optimal Version Updates
Gad Allon, Northwestern University,
g-allon@kellogg.northwestern.eduMobile apps have become an economy with a market size of $25 Billion in 2013
and with a projected market size of $77 Billion by 2017. One of the key features
that distinguishes mobile apps from other types of digital goods (such as movies,
songs or books) is the they have versions. A developer can release an app into a
mobile app store, and can then keep adding, removing or editing features of the
app with subsequent version updates. We study empirically and theoretically the
optimal strategy for such updates.
TA33